Re: testing for independence of 2 ordinal scales

On 26 Jul, 21:21, Paul Rubin <ru...@xxxxxxx> wrote:
claire wrote:

The 246 speech items comprise: 6 speakers each saying the same 41
items.
The speakers are:

2 native English, noimpediment,
1 native English, withimpediment,
1 native French, noimpediment,
1 native Italian, noimpediment,
1 native Greek, noimpediment.

The presence of only one speaker with animpedimentmay be damaging --
recognition of theimpedimentcould be confounded with other
characteristics of the speaker.  For instance, if the speaker has animpedimentand a regional accent,impedimentsounds might be perceived
as accent.

The fact that each speaker read the same 41 scripts is an issue,
although I don't know ultimately whether it's an advantage or
disadvantage.  It creates two correlations among the observations for
any given listener.  Obviously, reactions to two scripts from the same
speaker can easily be correlated (in fact, one might expect that).
Perhaps a bit less obviously, reactions to different speakers reading
the same script might be correlated (again, within one listener), since
a particular script might be more difficult to pronounce, involve words
less familiar to the listener, emphasize an accent issue (a script with,
say, a lot of hard "c"/"k" sounds might make certain accents more
noticeable).

I'm not sure if a chi-square test of independence of the two variables
would work (I'm not sure if there's a way to introduce "blocking
factors"), and while a paired-difference t-test benefits from having the
two variables measured on the same subjects, I think there's an
assumption that the differences all come from one population.  Even
within listeners, I'm not sure that holds up here, given multiple
observations from the same speaker and multiple observations from the
same script.

I wish I had something constructive to suggest, but I'm really not sure
what a rigorous way to proceed would be, given all this dependence among
observations.

/Paul

Thanks for your time. I'm reassured at least that I wasn't missing
something obvious I should be doing! Not quite sure how I'm going to
proceed but I have more than enough other analysis to do to keep me
busy for the moment!

Thanks again for your input, everyone.

Claire
.

Relevant Pages

  • background process/command in a script
    ... otherwise the listener just hangs. ... My issue is that I have to put this in as the startup script for a VCS ... echo "Starting MQ Listener for $on port ...
    (SunManagers)
  • Re: Win32::Process - Limiting the number of processes launched from within a script
    ... > on the script server. ... I might try writing a listener that I could put ... on every end point that listened on say port 9010 (or some other ... some other port) to handle replies to requests sent to the listener port ...
    (comp.lang.perl.misc)
  • Re: Problem with automatic startup
    ... I have a problem with database automatic startup. ... exactly database but listener. ... I've created dbora script that looks like this: ...
    (comp.databases.oracle.server)
  • Re: vwait
    ... listener running as the rest of my script is running. ... run scripts that will send messages to socket listener from multiple ...
    (comp.lang.tcl)